Hammerspace Storage Management MCP Server

# Web UI Testing Guide ## Quick Start ### Start the Web UI ```bash cd /home/mike/mcp-1.5/web_ui source ../venv/bin/activate python app.py ``` ### Access the UI Open your browser: http://localhost:5000 ## Running Tests ### Option 1: Curl Script (Fast, No Dependencies) ```bash cd /home/mike/mcp-1.5 ./tests/test_web_ui_curl.sh ``` ### Option 2: Python Test Suite (More Detailed) ```bash cd /home/mike/mcp-1.5 source venv/bin/activate python tests/test_web_ui.py ``` ### Option 3: Single Question Test ```bash ./tests/test_single_question.sh ``` ## Important Notes ### Conversation State The web UI currently starts each conversation fresh to avoid Claude API tool_result errors. This means: - ✅ Each question is independent and works reliably - ⚠️ No conversation context between questions - 💡 Click "Clear Chat" in the UI between complex multi-step operations ### Test Results Summary **Working Features:** - ✅ Server health check - ✅ MCP tools endpoint (19 tools) - ✅ Single question processing - ✅ Tool execution and results **Expected Behavior:** - Each question should be asked independently - Click "Clear Chat" between questions for best results - Responses appear in 5-15 seconds depending on complexity ## Example Test Flow ```bash # 1. Clear chat curl -X POST http://localhost:5000/api/clear # 2. Ask a question curl -X POST http://localhost:5000/api/chat \ -H "Content-Type: application/json" \ -d '{"message": "List all objectives for /mnt/se-lab/modelstore/gtc-demo-models/"}' # 3. Clear chat before next question curl -X POST http://localhost:5000/api/clear # 4. Ask another question curl -X POST http://localhost:5000/api/chat \ -H "Content-Type: application/json" \ -d '{"message": "Check alignment status of files tagged with modelsetid=hs-GTC-0002"}' ``` ## Troubleshooting ### Server Won't Start ```bash # Check if port 5000 is in use lsof -i :5000 # Kill existing processes pkill -f "python.*app.py" # Check logs tail -f /home/mike/mcp-1.5/logs/web_ui.log ``` ### API Errors ```bash # Verify Anthropic API key grep ANTHROPIC_API_KEY /home/mike/mcp-1.5/.env # Test API connectivity curl -I https://api.anthropic.com/v1/messages ``` ### MCP Tools Not Working ```bash # Test MCP server directly cd /home/mike/mcp-1.5 source venv/bin/activate python tmp/promote_gtc_to_tier0_mcp.py ``` ## Performance Tips 1. **Response Time**: 5-15 seconds per question (includes Claude AI + MCP tool execution) 2. **Clear Chat**: Always clear between independent questions 3. **Complex Operations**: Break into smaller steps if needed 4. **Network**: Ensure stable connection to api.anthropic.com ## Known Limitations 1. **No Conversation Memory**: Each question is independent 2. **Model Deprecation Warning**: Claude 3.5 Sonnet 20241022 is deprecated (update to newer model) 3. **Single Tool Per Question**: Best results with one action per question ## Future Enhancements - [ ] Proper conversation state management - [ ] Session-based chat history - [ ] Multi-step workflow support - [ ] Upgrade to latest Claude model - [ ] Add authentication - [ ] Rate limiting - [ ] WebSocket for real-time updates